Accuracy evaluation of commercial deformation registration systems using public datasets(PDF)
《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]
- Issue:
- 2026年第7期
- Page:
- 898-904
- Research Field:
- 医学影像物理
- Publishing date:
Info
- Title:
- Accuracy evaluation of commercial deformation registration systems using public datasets
- Author(s):
- LI Changhu; LI Xia; FU Yuchuan; ZENG Xianhu
- Radiotherapy Physics and Technology Center/Cancer Center, West China Hospital, Sichuan University, Chengdu 610041, China
- Keywords:
- Keywords: deformable image registration adaptive radiotherapy four-dimensiond computed tomography target registration error quality control
- PACS:
- R318；R811.1
- DOI:
- DOI:10.3969/j.issn.1005-202X.2026.07.009
- Abstract:
- Abstract: Objective To evaluate the accuracy of various deformable image registration (DIR) algorithms integrated in commercial systems for thoracic image registration using publicly available datasets. Methods Permission to utilize the 4D-CT image dataset containing 10 thoracic cancer patients, provided by the DIR Lab at Emory University School of Medicine (www.DIR-lab.com), was obtained after official registration. This dataset included anatomical landmarks identified and registered by thoracic imaging experts, together with detailed coordinate lists. Precise point-to-point correspondences between end-inspiratory and end-expiratory phase images were established to evaluate the registration accuracy of different DIR algorithms integrated in the Raystation treatment planning system and MIM software. Results For the RayStation system, the target registration error ranges of the evaluated algorithms were as follows: Biomechanical (0.20-30.50 mm), Control POI (0.00-28.20 mm), Control ROI (0.10-31.70 mm), and Focus ROI (0.00-31.00 mm). For the image based registration algorithm in the MIM system, the target registration error range was 0.09-32.45 mm. The error distribution across patients and registration modes and statistical analysis demonstrated significant inter-patient variability for each registration algorithm. Notably, different DIR algorithms produced markedly varying registration outcomes. Conclusion: The proposed evaluation method enables quantitative verification of the geometric accuracy of DIR algorithms embedded in commercial systems for clinical practice.
Last Update: 2026-07-22